The algorithm knows you — Australia wants the power to switch it off

Australia is preparing to confront one of the most powerful and least visible forces influencing modern life: the social media algorithm.

Proposed federal laws would require major online platforms to give Australians the ability to opt out of algorithmically selected content. Companies could also face penalties exceeding $100 million if they fail to meet expanded obligations to protect users from foreseeable harm.

The proposal is not yet law, and important details remain subject to consultation and parliamentary debate. Nevertheless, its central principle is significant.

Australians may be given the right to tell social media companies: stop deciding what I should see next.

That could represent a fundamental shift in the relationship between users and platforms.

For years, Australians have been able to choose which social media services they join, which people they follow and what content they post. What they have not truly controlled is the stream of material placed before them.

The platform makes that decision.

The invisible editor in every telephone

A social media feed may appear to be a simple collection of posts, videos, advertisements and news from accounts followed by the user.

It is usually nothing of the sort.

Recommendation systems study almost every available signal: which posts a person opens, how long they watch, what they share, what they ignore, where they pause and what causes them to return.

Even failing to press the “like” button does not necessarily conceal a preference. Lingering over a video can be enough to tell the system that the subject held the user’s attention.

The algorithm then predicts what is most likely to keep that person watching.

This can be convenient. Someone interested in gardening, photography, travel or restoring old cars may discover useful material that would otherwise have remained hidden.

But the system is not necessarily designed to determine what is accurate, constructive or good for the user. Its commercial objective is generally to maintain attention and encourage repeated engagement.

Content that produces anger, fear, anxiety or outrage can be particularly effective because emotionally charged users often keep watching, commenting and returning.

One controversial video can lead to another. A passing curiosity can become an apparently endless stream.

The user may believe they are exploring the internet. In reality, the internet is progressively narrowing itself around the user.

What the proposed laws would do

The Albanese Government is developing a broader “digital duty of care” intended to place greater responsibility on online platforms.

Under the proposed approach, users would be offered a way to opt out of algorithmically recommended feeds. Platforms would also be expected to identify and reduce foreseeable risks created by their services, particularly risks affecting children and vulnerable users.

Additional categories of potential psychological or social harm for younger people could cover matters such as bullying, harmful body-image material and content that repeatedly promotes violence or misogyny.

The proposed penalties are deliberately substantial. Large international technology businesses earn billions of dollars and operate across numerous jurisdictions. A small regulatory fine can simply become another cost of doing business.

A penalty exceeding $100 million is intended to be noticed in the boardroom.

The eSafety Commissioner has already examined the effects of recommender systems and argues that online services should be designed with safety in mind. The regulator’s position is that platforms should take reasonable steps to prevent foreseeable harm rather than waiting until damage has occurred.

That is the digital equivalent of obligations already accepted in many physical industries.

A manufacturer cannot knowingly sell a dangerous product and escape responsibility merely because the customer chose to purchase it. A workplace cannot ignore an obvious hazard and argue that employees voluntarily entered the building.

The government’s case is that technology companies should not be exempt from a comparable duty simply because their products exist on screens.

What would an algorithm-free feed look like?

This is one of the most important practical questions.

Switching off recommendations would not necessarily mean abandoning social media. It could mean returning to a chronological feed containing posts from people and organisations the user has deliberately chosen to follow.

Instead of the platform continually introducing provocative or addictive material, the user would regain greater control over what appears.

There would still be advertising. There would still be misleading claims, unpleasant opinions and poor-quality content. An opt-out button cannot make the internet harmless.

It may, however, reduce the ability of a platform to amplify material merely because its systems predict that it will provoke a reaction.

The value of the reform will depend heavily on its design.

An opt-out control buried inside several layers of settings would technically provide a choice while ensuring that few people exercised it. A clear and persistent option would be more meaningful.

There is also an important distinction between opting out and opting in.

Under an opt-out system, algorithmic recommendations remain the default unless the user takes action. Under an opt-in model, the platform would need permission before activating them.

That difference will be central to the debate.

Protecting users or controlling information?

The proposal inevitably raises concerns about censorship and government influence.

Those concerns should not be dismissed.

Any law governing online content must define harm carefully. A government should not be given broad or ambiguous authority to suppress lawful political opinions, unpopular arguments or legitimate reporting merely because someone considers them distressing.

Adults have the right to encounter ideas with which they disagree. Public debate will become poorer if “safety” is expanded until it means protection from discomfort, criticism or dissent.

The government must therefore demonstrate that its proposed duty applies to the design and conduct of platforms rather than becoming a mechanism for controlling lawful speech.

There is a meaningful distinction between directing a company to reduce addictive or manipulative systems and directing it to suppress a political viewpoint.

The first addresses how information is promoted. The second determines which ideas may be heard.

Australia must not allow that boundary to become blurred.

The opposition has expressed scepticism about the proposal and warned of the potential for censorship. That scrutiny is warranted, but simply leaving the matter to parents and individual users is not a complete response.

Parents cannot supervise every minute of a teenager’s online activity. Nor can an ordinary user independently examine an algorithm protected by commercial secrecy and altered continuously by one of the world’s largest corporations.

Personal responsibility matters, but it cannot operate effectively when one party does not understand the system and the other designed it.

Australia is becoming an international testing ground

Australia has already taken an unusually assertive position on online safety, including minimum-age restrictions for social media accounts.

The proposed algorithm reforms would move beyond the question of whether children should use certain services. They address the deeper issue of what those services do to every person once access is granted.

Other governments will be watching.

If Australia can create a workable right to reject algorithmic manipulation without suppressing lawful expression, it could offer a model for other countries.

If the legislation is vague, technically ineffective or easily exploited for political purposes, it could instead demonstrate the dangers of regulating a system few legislators fully understand.

That is why the parliamentary process matters.

Lawmakers should require clear definitions, transparent enforcement, independent oversight and regular public reporting. Researchers must also be able to examine whether the measures are producing genuine improvements rather than merely encouraging platforms to redesign their menus and declare compliance.

The larger question is who remains in control

Social media platforms did not become enormously valuable simply by allowing people to communicate.

Their greater achievement was learning how to predict and influence human attention at an extraordinary scale.

The product is not merely the website or application. It is the sequence of material shown to each individual and the behavioural response that sequence produces.

Australians should understand that an algorithm does not truly “know” them in the human sense. It recognises patterns, categorises behaviour and makes predictions.

Those predictions can be remarkably accurate. They can also gradually shape the very preferences they claim merely to observe.

Giving users an off switch will not resolve every problem associated with social media. It may nevertheless establish an important principle: participation in modern digital life should not require surrendering control over the information placed before us.

The Times View

Australians deserve greater control over their social media feeds, and technology companies should bear responsibility for foreseeable harm created by systems they design and profit from.

The proposed digital duty of care is therefore worthy of serious consideration.

But protecting Australians must not become an excuse for deciding which lawful opinions Australians may encounter. The final legislation should regulate manipulative platform design, demand transparency and protect children while preserving free expression.

The best outcome is neither an unregulated digital marketplace nor a government-managed public conversation.

It is an internet in which Australians can make informed choices—and in which the algorithm serves the user rather than quietly controlling them.